An ultrasonic image processing method for swallowing function examination

Abstract

The present invention discloses an ultrasonic image processing method for swallowing function examination, mainly relating to the technical field of ultrasonic image processing. It includes: arranging an ultrasonic probe according to a preset placement position and placement angle to capture a sequence of dynamic ultrasonic image frames; obtaining an enhanced sequence of dynamic ultrasonic image frames; obtaining a set sequence of watershed lines; obtaining a sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasonic image frames; extracting the hyoid-thyroid cartilage distance to obtain a set of hyoid-thyroid cartilage distances; obtaining the shortening rate of the target hyoid-thyroid cartilage distance, and taking the shortening rate of the target hyoid-thyroid cartilage distance as the ultrasonic image processing result. The present invention solves the technical problems of poor quality of ultrasonic image processing for swallowing function examination and low reliability of processing results in the prior art. The beneficial effects of the present invention are as follows: improving the reliability of ultrasonic image processing and enhancing the recognition accuracy of the shortening rate of the hyoid-thyroid cartilage distance.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic image processing, and particularly relates to an ultrasonic image processing method for swallowing function examination. Background Art

[0002] The swallowing function examination mainly observes and evaluates the swallowing process through imaging techniques to judge the patient's swallowing ability and possible disorders. As a commonly used imaging examination method, ultrasonic images have the advantages of being non-invasive, real-time, and convenient, and can reflect the swallowing movements in the oral and laryngeal phases. In ultrasonic images, the positional relationship between the hyoid bone and the thyroid cartilage is an important basis for judging swallowing function. However, existing ultrasonic image processing methods usually use simple image filtering or contrast adjustment to improve image quality, and cannot effectively remove noise, resulting in blurred image details. In particular, the minute movements during swallowing cannot be clearly captured.

[0003] There are technical problems in the prior art such as poor quality of ultrasonic image processing for swallowing function examination and low reliability of processing results. Summary of the Invention

[0004] The present application provides an ultrasonic image processing method for swallowing function examination, aiming to solve the technical problems of poor quality of ultrasonic image processing for swallowing function examination and low reliability of processing results in the prior art.

[0005] In view of the above problems, the present application provides an ultrasonic image processing method for swallowing function examination, and the method includes:

[0006] Arrange the ultrasonic probe according to a preset placement position and placement angle, and use the arranged ultrasonic probe to capture the dynamic ultrasonic image of the target examination object swallowing a preset milliliter of water at one time, and obtain a dynamic ultrasonic image frame sequence;

[0007] Introduce a noise function, and perform image enhancement processing on the dynamic ultrasonic image frame sequence in combination with a pre-trained denoising function to obtain an enhanced dynamic ultrasonic image frame sequence;

[0008] Simulate the pixel value of each pixel point as the height of a simulated hillside, traverse the enhanced dynamic ultrasonic image frame sequence for hillside simulation, and obtain a simulated hillside sequence;

[0009] Respectively take the position where the minimum hillside height of the simulated hillside sequence is located as the water injection point, inject water into the simulated hillside sequence, and identify the dividing ridge line of the simulated hillside sequence according to a preset hillside height difference threshold as the water surface rises, and obtain a dividing ridge line set sequence;

[0010] Based on the positions of the dividing ridgeline in the sequence of dividing ridgeline sets, perform neighborhood contour recognition on the sequence of enhanced dynamic ultrasound image frames corresponding to the simulated slope sequence to obtain a sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames;

[0011] Traverse the sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames to extract the hyoid-thyroid cartilage distance, and obtain a set of hyoid-thyroid cartilage distances;

[0012] Based on the magnitudes of the hyoid-thyroid cartilage distances in the set of hyoid-thyroid cartilage distances, perform recognition of the shortening rate of the hyoid-thyroid cartilage distance to obtain the target shortening rate of the hyoid-thyroid cartilage distance, and use the target shortening rate of the hyoid-thyroid cartilage distance as the result of ultrasound image processing.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages: In this application, the ultrasound probe is arranged according to the preset placement position and placement angle, and the dynamic ultrasound image of the target examination object swallowing a preset milliliter of water at one time is captured by the arranged ultrasound probe to obtain a sequence of dynamic ultrasound image frames. Then, a noise function is introduced, and the sequence of dynamic ultrasound image frames is subjected to image enhancement processing in combination with a pre-trained denoising function to obtain a sequence of enhanced dynamic ultrasound image frames. Furthermore, the pixel value of each pixel point is simulated as the height of a simulated hillside of a simulated slope, and the sequence of enhanced dynamic ultrasound image frames is traversed for slope simulation to obtain a sequence of simulated slopes. The positions where the minimum hillside heights of the sequence of simulated slopes are located are respectively used as water injection points, and water is injected into the sequence of simulated slopes. As the water surface rises, the dividing ridgelines are recognized for the sequence of simulated slopes according to the preset threshold of the hillside height difference to obtain a sequence of dividing ridgeline sets. Then, based on the positions of the dividing ridgelines in the sequence of dividing ridgeline sets, neighborhood contour recognition is performed on the sequence of enhanced dynamic ultrasound image frames corresponding to the sequence of simulated slopes to obtain a sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames. The sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames is traversed to extract the hyoid-thyroid cartilage distance, and a set of hyoid-thyroid cartilage distances is obtained. Based on the magnitudes of the hyoid-thyroid cartilage distances in the set of hyoid-thyroid cartilage distances, recognition of the shortening rate of the hyoid-thyroid cartilage distance is performed to obtain the target shortening rate of the hyoid-thyroid cartilage distance, and the target shortening rate of the hyoid-thyroid cartilage distance is used as the result of ultrasound image processing. It achieves the technical effects of significantly improving the reliability of ultrasound image processing in swallowing function examination and improving the measurement accuracy of the hyoid-thyroid cartilage distance. Brief Description of the Drawings

[0014] Attached Figure 1 is a schematic flowchart of a method for ultrasound image processing for swallowing function examination provided by an embodiment of the present invention.

[0015] Attached Figure 2It is a schematic flowchart of obtaining a set of watershed line sequences in an ultrasonic image processing method for swallowing function examination provided by an embodiment of the present invention. Detailed implementation manners

[0016] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of this application.

[0017] Embodiment, as Figure 1 shown, the present application provides an ultrasonic image processing method for swallowing function examination, wherein the method includes:

[0018] S100: Deploy an ultrasonic probe according to a preset placement position and placement angle, and use the deployed ultrasonic probe to capture the dynamic ultrasonic images of a target examination object swallowing a preset milliliter of water at one time, and obtain a sequence of dynamic ultrasonic image frames;

[0019] In an embodiment of the present application, the ultrasonic probe is placed in the throat area of the patient according to the preset placement position and placement angle set by those skilled in the art according to the actual situation of the target monitoring object, so as to clearly capture the process of the swallowing action. When the target object (the subject) swallows a preset milliliter (such as 5 milliliters) of water at one time, real-time ultrasonic image acquisition of the swallowing action is performed through an ultrasonic device, and the dynamic ultrasonic image sequence is obtained. Among them, the dynamic ultrasonic image sequence refers to a series of continuous image frames collected by the ultrasonic probe, and each frame represents a certain moment in the swallowing process. Since the dynamic ultrasonic image sequence is continuous, the rapid changes occurring during the swallowing process can be captured.

[0020] By obtaining the sequence of dynamic ultrasonic image frames, the technical effects of providing raw data for subsequent image enhancement, noise removal, and structural analysis, capturing subtle physiological changes during swallowing, and providing basic data for the measurement of the hyoid-thyroid cartilage distance are achieved.

[0021] S200: Introduce a noise function, and perform image enhancement processing on the sequence of dynamic ultrasonic image frames in combination with a pre-trained denoising function to obtain an enhanced sequence of dynamic ultrasonic image frames;

[0022] Furthermore, introduce a noise function, and perform image enhancement processing on the sequence of dynamic ultrasonic image frames in combination with a pre-trained denoising function to obtain an enhanced sequence of dynamic ultrasonic image frames. Step S200 of the embodiment of the present application further includes:

[0023] Obtain a noise function, wherein the noise function is:

[0024] x i =α t y i +(1-α t )·∈ t ;

[0025] Among them, x i is the i-th training sample noise ultrasound image frame in the training sample noise ultrasound image frame set, y i is the i-th training sample ultrasound image frame in the training sample ultrasound image frame set, α t is the preset diffusion intensity, ∈ t is the preset random noise;

[0026] The dynamic ultrasonic image frame sequence is forward diffused based on the noise function, and the forward diffused dynamic ultrasonic image frame sequence is reverse diffused using the pre-trained denoising function to obtain the enhanced dynamic ultrasonic image frame sequence.

[0027] Furthermore, step S200 in the embodiment of the present application further includes:

[0028] Acquire a training sample initial ultrasound image frame set, perform noise diffusion processing on the sample ultrasound image frame set based on the noise function, and obtain a training sample noise ultrasound image frame set, wherein the training sample ultrasound image frames and the training sample noise ultrasound image frames have a one-to-one correspondence;

[0029] Obtaining images of the initial ultrasound image frame set of the training sample after image enhancement to obtain an enhanced ultrasound image frame set of the training sample;

[0030] The coefficients of the pre-constructed denoising function are fitted and solved based on the training sample ultrasound image frame set, the noise function, the training sample noise ultrasound image frame set and the training sample enhanced ultrasound image frame set, and the pre-trained denoising function is obtained according to the solution result.

[0031] Furthermore, step S200 in the embodiment of the present application further includes:

[0032] Obtain a pre-built denoising function, wherein the pre-built denoising function is:

[0033]

[0034] in, For the i-th training sample enhanced ultrasound image frame in the training sample enhanced ultrasound image frame set, x i is the i-th training sample noise ultrasound image frame in the training sample noise ultrasound image frame set, y iis the i-th training sample ultrasound image frame in the set of training sample ultrasound image frames, σ t is a noise function, and ε is a denoising coefficient.

[0035] In a possible embodiment, the noise function is a mathematical function used to simulate the influence of noise in an image. The noise function is used to combine noise with the dynamic ultrasound image frame sequence to generate a noisy image, and then a pre-trained denoising function is used to perform image enhancement processing on the noisy image, so as to obtain a clearer and higher-quality enhanced dynamic ultrasound image frame sequence. This achieves the technical effect of providing data support for subsequent image contour recognition.

[0036] In a possible embodiment, the noise function is:

[0037] x i = α t y i +(1 - α t )·∈ t ;

[0038] where x i is the i-th training sample noisy ultrasound image frame in the set of training sample noisy ultrasound image frames, y i is the i-th training sample ultrasound image frame in the set of training sample ultrasound image frames, α t is a preset diffusion intensity, and ∈ t is a preset random noise.

[0039] Optionally, the preset random noise can be Gaussian noise, representing random perturbations in the image. The preset diffusion intensity is a diffusion intensity preset by those skilled in the art, used to control the intensity of the noise. Forward diffusion refers to the process of gradually converting a clear image into a noisy image by adding noise in image processing. In this step, the noise function is used to perform forward diffusion on the image to simulate the influence of noise. Reverse diffusion is a process opposite to forward diffusion, aiming to remove noise and restore the image to its original state. In this step, through the pre-trained denoising function, the forward diffusion process is reversed to recover a clear image from the noisy image. The pre-trained denoising function is an image processing function used to recover a clear image from a noisy image. It can effectively process image noise by learning the features of noise removal on a large amount of training data.

[0040] In one embodiment, the set of initial ultrasound image frames of the training samples is a set containing real, noise-free ultrasound image frames, which are used to train the denoising model during the training process. The set of enhanced ultrasound image frames of the training samples is obtained by a person skilled in the art after image enhancement based on the set of initial ultrasound image frames of the training samples. The set of noisy ultrasound image frames of the training samples is obtained by adding noise to the image frames in the set of sample ultrasound image frames based on the noise function.

[0041] Obtain a pre-constructed denoising function, and the pre-constructed denoising function is:

[0042]

[0043] Where is the i-th enhanced ultrasound image frame of the training samples in the set of enhanced ultrasound image frames of the training samples, x i is the i-th noisy ultrasound image frame of the training samples in the set of noisy ultrasound image frames of the training samples, y i is the i-th ultrasound image frame of the training samples in the set of ultrasound image frames of the training samples, σ t is the noise function, and ε is the denoising coefficient.

[0044] Optionally, the denoising coefficient is used to adjust the denoising effect. is the square of the Euclidean distance, representing the degree of difference between the noisy ultrasound image frame of the training samples and the ultrasound image frame of the training samples.

[0045] Input the set of ultrasound image frames of the training samples, the noise function, the set of noisy ultrasound image frames of the training samples, and the set of enhanced ultrasound image frames of the training samples into the pre-constructed denoising function for fitting analysis to obtain the solved denoising coefficient. According to the solved denoising coefficient, obtain the pre-trained denoising function that has been trained. Through training, optimize the combination of the noise function and the denoising function to improve the quality of image enhancement.

[0046] S300: Simulate the pixel value of each pixel point as the height of a simulated hillside, traverse the sequence of enhanced dynamic ultrasound image frames for hillside simulation, and obtain a sequence of simulated hillsides;

[0047] S400: Respectively use the position where the minimum hillside height of the sequence of simulated hillsides is located as the water injection point, inject water into the sequence of simulated hillsides, and identify the dividing ridgeline of the sequence of simulated hillsides according to the preset hillside height difference threshold as the water surface rises, and obtain a sequence of sets of dividing ridgelines;

[0048] Further, as shown in the appendix Figure 2As shown, the positions where the minimum hillside height of the simulated hillside sequence is located are respectively used as water injection points, and water is injected into the simulated hillside sequence. As the water surface rises, ridge lines are identified for the simulated hillside sequence according to a preset hillside height difference threshold, and a set sequence of ridge lines is obtained. Step S400 of the embodiment of the present application further includes:

[0049] Extract the first simulated hillside from the simulated hillside sequence, and use the minimum hillside height of the first simulated hillside as the benchmark for threshold judgment;

[0050] Based on the water injection point of the first simulated hillside, water is injected. As the water surface rises, when the water surface submerges a simulated hillside, the first water surface submerging the simulated hillside is obtained;

[0051] Judge whether the hillside height difference between the first water surface submerging the simulated hillside and the hillside height of the water injection point of the first simulated hillside is greater than or equal to the preset hillside height difference threshold. If not, continue to inject water, and identify the ridge line based on the minimum hillside height and the preset hillside height difference threshold;

[0052] If so, generate a ridge line at the position where the first water surface submerges the simulated hillside, continue to inject water, and obtain the first set of ridge lines. Among them, the ridge line rises as the water surface rises;

[0053] The positions where the minimum hillside height of the simulated hillside sequence is located are respectively used as water injection points, and water is injected into the simulated hillside sequence. During the rising process of the water surface, ridge lines are identified for the simulated hillside sequence according to the preset hillside height difference threshold, and a set sequence of ridge lines is obtained.

[0054] Further, if so, generate a ridge line at the position where the first water surface submerges the simulated hillside, continue to inject water, and obtain the first set of ridge lines. Step S400 of the embodiment of the present application further includes:

[0055] Update the hillside height of the first water surface submerging the simulated hillside as the benchmark for threshold judgment;

[0056] When the water surface submerges a simulated hillside again, the second water surface submerging the simulated hillside is obtained;

[0057] Judge whether the height difference between the hillside height of the second water surface submerging the simulated hillside and the hillside height of the first water surface submerging the simulated hillside is greater than or equal to the preset hillside height difference threshold. If so, generate a ridge line at the position where the second water surface submerges the simulated hillside, and update the hillside height of the second water surface submerging the simulated hillside as the benchmark for threshold judgment, and continue to inject water for ridge line identification until the water surface submerges the highest point of the first simulated hillside, then stop injecting water, and obtain the first set of ridge lines.

[0058] In a possible embodiment, the pixel value of each pixel is associated with the height of a hillside to form a model analogous to a hillside. The simulated hillside height reflects the intensity of the pixel values in the image (such as grayscale values), where higher pixel values may represent higher hillsides and lower pixel values represent lower hillsides. By analogy between the pixel value of each pixel and the hillside height, the enhanced dynamic ultrasound image frame sequence is transformed into a simulated slope sequence. This process simulates the pixel value of each pixel in the image as the "hillside" height, making the analysis of the entire image more structured and facilitating subsequent more complex analysis.

[0059] In the simulated hillside terrain, water injection is carried out by simulating the liquid rising from the lower part of the hillside, reflecting important regions or key points in the image. Preferably, the process of water injection is similar to "filling" the image, which can help identify different regions. In the simulated slope, the dividing ridgeline is the boundary defined by the water injection process and the rising water level. These dividing lines help divide different regions, thereby achieving the recognition and analysis of the image structure. The preset hillside height difference threshold is the maximum hillside height difference that can be divided into one region preset by those skilled in the art. By obtaining the sequence of dividing ridgeline sets, the technical effect of laying a foundation for subsequent neighborhood contour recognition of the enhanced dynamic ultrasound image frame sequence is achieved.

[0060] Preferably, a first simulated slope is extracted from the simulated slope sequence, and the minimum hillside height of the first simulated slope is used as the benchmark for threshold judgment. The first simulated slope is a sub-region extracted from the simulated slope sequence for further water injection and hillside height difference judgment.

[0061] Water is injected into the slope from the water injection point of the first simulated slope, simulating the liquid rising from the lower part of the hillside. As the water surface rises and passes over a simulated hillside, the first water surface passing over the simulated hillside is obtained. Furthermore, it is judged whether the hillside height difference between the first water surface passing over the simulated hillside and the water injection point of the first simulated slope is greater than or equal to the preset hillside height difference threshold. If not, it indicates that the difference between the first simulated slope and the first water surface passing over the simulated hillside is not significant and can be divided into one neighborhood without ridge line separation. Continue to inject water and identify the dividing ridgeline based on the minimum hillside height and the preset hillside height difference threshold.

[0062] If so, it indicates that there is a significant difference between the first simulated slope and the first water surface submerging the simulated hillside. It is necessary to generate a dividing ridge line at the position where the first water surface submerges the simulated hillside, continue to inject water, and update the hillside height where the first water surface submerges the simulated hillside as the benchmark for threshold judgment. When the water surface submerges a simulated slope again, the second water surface submerging the simulated hillside is obtained. Further, it is judged whether the height difference between the hillside height where the second water surface submerges the simulated hillside and the hillside height where the first water surface submerges the simulated hillside is greater than or equal to the preset hillside height difference threshold. If so, a dividing ridge line is generated at the position where the second water surface submerges the simulated hillside, and the hillside height where the second water surface submerges the simulated hillside is updated as the benchmark for threshold judgment, and continue to inject water for dividing ridge line identification until the water surface submerges the highest point of the first simulated slope, then stop injecting water, and summarize the dividing ridge lines obtained during the water injection process to obtain the first dividing ridge line set. Among them, the dividing ridge line rises as the water surface rises.

[0063] Based on the same principle as obtaining the first dividing ridge line set, the position of the minimum hillside height of the simulated slope sequence is respectively used as the water injection point, water is injected into the simulated slope sequence, and dividing ridge line identification is performed on the simulated slope sequence based on the preset hillside height difference threshold during the rising process of the water surface, and a dividing ridge line set sequence is obtained.

[0064] By respectively taking the minimum hillside height of the simulated slope sequence as the water injection point, start to "inject water" into the slope. As the water surface rises, according to the set hillside height difference threshold, the dividing ridge line (similar to the regional boundary in the image) is identified. This process simulates how to identify the regions with key features in the image. In this way, the system can gradually identify and separate different regions by the rising of the water level, and then help to extract important information in the image, such as the structural features like the distance between the hyoid bone and the thyroid cartilage.

[0065] This method effectively transforms the image segmentation problem into a terrain analysis problem, uses the "water injection" process to identify the key regions in the image through simulating the hillside, and then improves the accuracy and reliability of image processing in swallowing function examination.

[0066] S500: Based on the positions of the dividing ridge lines in the dividing ridge line set sequence, perform neighborhood contour recognition on the sequence of enhanced dynamic ultrasound image frames corresponding to the simulated slope sequence, and obtain a sequence of enhanced dynamic ultrasound image frame neighborhood contour recognition results;

[0067] In a possible embodiment, the important contours (such as the edges of objects) in the image are identified and extracted by analyzing the relationship between adjacent pixels in the image to perform neighborhood contour recognition. In step S500, the neighborhood contour recognition is based on the positions of the dividing ridge lines, and the regional contours containing key information are respectively identified in the enhanced dynamic ultrasound image frames.

[0068] Preferably, according to the positions of the dividing ridge lines in the sequence of dividing ridge line sets, the dividing ridge lines in a dividing ridge line set are connected to each other, and the contour of the neighborhood area is extracted, so as to obtain neighborhoods of different structures (for example, the neighborhood of the hyoid bone or the thyroid cartilage), for further analysis. It achieves the technical effect of providing support for the subsequent measurement of the distance between the hyoid bone and the thyroid cartilage.

[0069] S600: Traverse the sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames to extract the distance between the hyoid bone and the thyroid cartilage, and obtain a set of distances between the hyoid bone and the thyroid cartilage;

[0070] Furthermore, traverse the sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames to extract the distance between the hyoid bone and the thyroid cartilage, and obtain a set of distances between the hyoid bone and the thyroid cartilage. Step S600 of the embodiment of the present application further includes:

[0071] Obtain the recognition results of the neighborhood contours of multiple sample enhanced dynamic ultrasound image frames, and perform manual labeling based on the recognition results of the neighborhood contours of the multiple sample enhanced dynamic ultrasound image frames to obtain multiple sample distances between the hyoid bone and the thyroid cartilage, where each distance between the hyoid bone and the thyroid cartilage corresponds to a recognition result of the neighborhood contour of a sample enhanced dynamic ultrasound image frame;

[0072] Use the recognition results of the neighborhood contours of the multiple sample enhanced dynamic ultrasound image frames and the multiple sample distances between the hyoid bone and the thyroid cartilage to supervise and train the framework constructed based on the convolutional neural network, learn the one-to-one mapping relationship between the recognition result of the neighborhood contour of the sample enhanced dynamic ultrasound image frame and the distance between the hyoid bone and the thyroid cartilage, until the training converges, and obtain a distance extractor;

[0073] Use the distance extractor to extract the distance between the hyoid bone and the thyroid cartilage from the sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames, and obtain the set of distances between the hyoid bone and the thyroid cartilage.

[0074] In the embodiment of the present application, the distance between the hyoid bone and the thyroid cartilage refers to the distance between the hyoid bone and the thyroid cartilage. In the examination of swallowing function, the change of this distance is of great significance for evaluating the functional state of the swallowing process. By measuring this distance, it can be evaluated whether the swallowing action is normal and whether the hyoid bone and the thyroid cartilage have appropriate displacements.

[0075] The sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames is the contour information (such as the hyoid bone and thyroid cartilage regions) of each key region in the enhanced dynamic ultrasound image frame sequence obtained in the previous steps.

[0076] Obtain the recognition results of the neighborhood contours of multiple sample enhanced dynamic ultrasound image frames. Manually mark the hyoid-thyroid cartilage distance in each image frame by an expert or operator to obtain multiple sample hyoid-thyroid cartilage distances. This process is to generate training data for subsequent training of machine learning algorithms. Among them, each hyoid-thyroid cartilage distance corresponds to a recognition result of the neighborhood contours of a sample enhanced dynamic ultrasound image frame.

[0077] Use the recognition results of the neighborhood contours of the multiple sample enhanced dynamic ultrasound image frames and the multiple sample hyoid-thyroid cartilage distances to perform supervised training on the framework constructed based on the convolutional neural network, and learn the one-to-one mapping relationship between the recognition results of the neighborhood contours of the sample enhanced dynamic ultrasound image frames and the hyoid-thyroid cartilage distances until the training converges to obtain a distance extractor. Among them, the convolutional neural network (CNN) is a deep learning model widely used in image processing and recognition tasks. Here, CNN is used to learn the relationship between the enhanced dynamic ultrasound image frames in the samples and the corresponding hyoid-thyroid cartilage distances. The distance extractor is a model obtained through CNN training and is used to automatically extract the hyoid-thyroid cartilage distance from the recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames.

[0078] Through the supervised training of CNN, the model (distance extractor) can learn to automatically extract the hyoid-thyroid cartilage distance from the image. This process is the core part of automatic recognition, which greatly improves the efficiency and accuracy of swallowing function examination. Finally, using the trained distance extractor, the hyoid-thyroid cartilage distance can be automatically extracted from the new sequence of recognition results of the neighborhood contours of the enhanced dynamic ultrasound image frames, thereby obtaining a set of hyoid-thyroid cartilage distances. The technical effect of improving the efficiency and accuracy of ultrasound image processing is achieved.

[0079] S700: Based on the size of the hyoid-thyroid cartilage distance in the set of hyoid-thyroid cartilage distances, perform recognition of the hyoid-thyroid cartilage distance shortening rate to obtain the target hyoid-thyroid cartilage distance shortening rate, and use the target hyoid-thyroid cartilage distance shortening rate as the ultrasound image processing result.

[0080] Furthermore, based on the size of the hyoid-thyroid cartilage distance in the set of hyoid-thyroid cartilage distances, perform recognition of the hyoid-thyroid cartilage distance shortening rate to obtain the target hyoid-thyroid cartilage distance shortening rate. Step S700 of the embodiment of the present application further includes:

[0081] Extract the maximum value and the minimum value of the hyoid-thyroid cartilage distance in the set of hyoid-thyroid cartilage distances;

[0082] Calculate the difference between the maximum value and the minimum value of the hyoid-thyroid cartilage distance, and use the ratio of the calculation result to the maximum value of the hyoid-thyroid cartilage distance as the target hyoid-thyroid cartilage distance shortening rate.

[0083] In one embodiment, the set of hyoid-thyroid cartilage distances refers to the set of distance values between the hyoid bone and the thyroid cartilage in all the image frames extracted through the foregoing steps. This set contains the actual distance data between the hyoid bone and the thyroid cartilage in each frame of the image. The maximum and minimum values of the hyoid-thyroid cartilage distance refer to the maximum and minimum distances among all the measured values in the set of hyoid-thyroid cartilage distances. Through these extreme values, the relative position changes of the hyoid bone and the thyroid cartilage during swallowing can be quantified. The target hyoid-thyroid cartilage distance shortening rate is an index to measure the relative distance change between the hyoid bone and the thyroid cartilage during the swallowing movement, and is specifically calculated as the ratio of the difference between the maximum distance and the minimum distance to the maximum distance. The larger this ratio is, the more significant the relative movement between the hyoid bone and the thyroid cartilage is, which helps to evaluate the effectiveness of the swallowing function.

[0084] By analyzing all the data in the set of hyoid-thyroid cartilage distances, first extract the maximum value and the minimum value, that is, the maximum and minimum distances between the hyoid bone and the thyroid cartilage. Then, calculate the difference between these two values and perform a ratio calculation with the maximum value to obtain the hyoid-thyroid cartilage distance shortening rate. This index is used to reflect the relative movement between the hyoid bone and the thyroid cartilage during swallowing, and further evaluate whether the swallowing function is normal. If this shortening rate is small, it may mean that there is an abnormality in the swallowing function, indicating that further examination or intervention is needed. Therefore, the hyoid-thyroid cartilage distance shortening rate, as the final result of ultrasonic image processing, is one of the key outputs of this method. It achieves the technical effect of improving the reliability of ultrasonic image processing.

[0085] In summary, the embodiments of the present application at least have the following technical effects:

[0086] Through image enhancement, the present application can effectively remove noise and blur, making the hyoid-thyroid cartilage distance during swallowing clearer and measurable. By simulating the slope and water injection operations, the pixel value of each image is analogized to the height of the hillside, and the structural changes in the image can be simulated at a finer granularity level, providing a good basis for subsequent ridge line recognition. Then, based on the position of the ridge line, through neighborhood contour recognition, the dynamic distance between the hyoid bone and the thyroid cartilage during swallowing can be accurately extracted, thereby improving the recognition accuracy of the distance shortening rate. It achieves the technical effect of significantly enhancing the reliability of ultrasonic image processing in swallowing function examination.

[0087] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. In addition, the specific embodiments of this specification have been described. Further, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0088] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0089] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. An ultrasonic image processing method for swallowing function examination, characterized in that: The method comprises: The ultrasound probe is deployed according to a preset placement position and placement angle, and a dynamic ultrasound image of the target inspection subject swallowing a preset milliliter of water at one time is captured using the deployed ultrasound probe to obtain a dynamic ultrasound image frame sequence; Introducing a noise function and performing image enhancement processing on the dynamic ultrasound image frame sequence in combination with a pre-trained denoising function to obtain an enhanced dynamic ultrasound image frame sequence; Simulating the pixel value of each pixel point as the height of a simulated hillside, traversing the enhanced dynamic ultrasound image frame sequence to perform slope simulation, and obtaining a simulated slope sequence; The locations of the minimum slope heights of the simulated slope sequence are respectively used as water injection points, and water is injected into the simulated slope sequence. As the water level rises, the dividing ridge lines of the simulated slope sequence are identified according to a preset slope height difference threshold, to obtain a dividing ridge line set sequence; Based on the position of the dividing ridge line in the dividing ridge line set sequence, performing neighborhood contour recognition on the enhanced dynamic ultrasound image frame sequence corresponding to the simulated slope sequence to obtain a neighborhood contour recognition result sequence of the enhanced dynamic ultrasound image frame; Traversing the enhanced dynamic ultrasound image frame neighborhood contour recognition result sequence to extract the hyoid bone-thyroid cartilage distance to obtain a hyoid bone-thyroid cartilage distance set; Based on the size of the hyoid bone-thyroid cartilage distance in the hyoid bone-thyroid cartilage distance set, identifying the hyoid bone-thyroid cartilage distance shortening rate, obtaining a target hyoid bone-thyroid cartilage distance shortening rate, and using the target hyoid bone-thyroid cartilage distance shortening rate as an ultrasound image processing result; The method of introducing a noise function and performing image enhancement processing on the dynamic ultrasound image frame sequence in combination with a pre-trained denoising function to obtain an enhanced dynamic ultrasound image frame sequence includes: Obtain a noise function, wherein the noise function is: ； in, is the i-th training sample noise ultrasound image frame in the training sample noise ultrasound image frame set, is the i-th training sample ultrasound image frame in the training sample ultrasound image frame set, is the preset diffusion intensity, is the preset random noise; The dynamic ultrasonic image frame sequence is forward diffused based on the noise function, and the forward diffused dynamic ultrasonic image frame sequence is reverse diffused using the pre-trained denoising function to obtain the enhanced dynamic ultrasonic image frame sequence.

2. The ultrasonic image processing method for swallowing function examination according to claim 1, characterized in that: include: Acquire a training sample initial ultrasound image frame set, perform noise diffusion processing on the sample ultrasound image frame set based on the noise function, and obtain a training sample noise ultrasound image frame set, wherein the training sample ultrasound image frames and the training sample noise ultrasound image frames have a one-to-one correspondence; Obtaining images of the initial ultrasound image frame set of the training sample after image enhancement to obtain an enhanced ultrasound image frame set of the training sample; The coefficients of the pre-constructed denoising function are fitted and solved based on the training sample ultrasound image frame set, the noise function, the training sample noise ultrasound image frame set and the training sample enhanced ultrasound image frame set, and the pre-trained denoising function is obtained according to the solution result.

3. The ultrasonic image processing method for swallowing function examination according to claim 2, characterized in that: include: Obtain a pre-built denoising function, wherein the pre-built denoising function is: ； in, For the i-th training sample enhanced ultrasound image frame in the training sample enhanced ultrasound image frame set, is the i-th training sample noise ultrasound image frame in the training sample noise ultrasound image frame set, is the i-th training sample ultrasound image frame in the training sample ultrasound image frame set, is the noise function, is the denoising coefficient.

4. The ultrasonic image processing method for swallowing function examination according to claim 1, characterized in that: The positions of the minimum hillside heights of the simulated slope sequence are respectively used as water injection points, and water is injected into the simulated slope sequence. As the water level rises, the dividing ridge lines of the simulated slope sequence are identified according to a preset hillside height difference threshold, and a dividing ridge line set sequence is obtained, including: Extracting a first simulated slope from the simulated slope sequence, and using the minimum slope height of the first simulated slope as a reference for threshold determination; Water is injected based on the water injection point of the first simulated slope, and as the water level rises, when the water surface covers a simulated slope, a first water surface covers the simulated slope is obtained; determining whether a difference in slope height between the simulated slope submerged by the first water surface and the water injection point of the first simulated slope is greater than or equal to a preset slope height difference threshold; if not, continuing water injection, and identifying a dividing ridge line based on the minimum slope height and the preset slope height difference threshold; If so, a dividing ridge line is generated at the point where the first water surface covers the simulated hillside, and water injection is continued to obtain a first dividing ridge line set, wherein the dividing ridge line rises as the water surface rises; The positions of the minimum hillside heights of the simulated slope sequence are respectively used as water injection points, and water is injected into the simulated slope sequence. During the rising process of the water level, the dividing ridge lines of the simulated slope sequence are identified based on the preset hillside height difference threshold to obtain a dividing ridge line set sequence.

5. The ultrasonic image processing method for swallowing function examination according to claim 4, characterized in that: If so, a dividing ridge line is generated where the first water surface covers the simulated hillside, and water injection is continued to obtain the first dividing ridge line set, including: The height of the simulated hillside at which the first water surface covers the hillside is updated as a benchmark for threshold judgment; When the water surface covers a simulated slope again, the second water surface covers the simulated slope; Determine whether the difference in height between the simulated hillside where the second water surface is submerged and the simulated hillside where the first water surface is submerged is greater than or equal to a preset hillside height difference threshold. If so, generate a dividing ridge line at the location where the second water surface is submerged in the simulated hillside, and update the height of the simulated hillside where the second water surface is submerged as the benchmark for threshold judgment. Continue to inject water to identify the dividing ridge line until the water surface is submerged in the highest point of the first simulated slope, then stop injecting water to obtain the first dividing ridge line set.

6. The ultrasonic image processing method for swallowing function examination according to claim 1, characterized in that: Traversing the enhanced dynamic ultrasound image frame neighborhood contour recognition result sequence to extract the hyoid bone-thyroid cartilage distance, and obtaining the hyoid bone-thyroid cartilage distance set, including: Acquiring multiple sample enhanced dynamic ultrasound image frame neighborhood contour recognition results, and manually marking the multiple sample enhanced dynamic ultrasound image frame neighborhood contour recognition results to obtain multiple sample hyoid bone-thyroid cartilage distances, wherein each hyoid bone-thyroid cartilage distance corresponds to one sample enhanced dynamic ultrasound image frame neighborhood contour recognition result; Using the multiple sample enhanced dynamic ultrasound image frame neighborhood contour recognition results and the multiple sample hyoid bone-thyroid cartilage distances to supervise the training of a framework built based on a convolutional neural network, learning a one-to-one mapping relationship between the sample enhanced dynamic ultrasound image frame neighborhood contour recognition results and the hyoid bone-thyroid cartilage distances until the training reaches convergence, thereby obtaining a distance extractor; The distance extractor is used to extract the hyoid bone-thyroid cartilage distance from the enhanced dynamic ultrasound image frame neighborhood contour recognition result sequence to obtain the hyoid bone-thyroid cartilage distance set.

7. The ultrasonic image processing method for swallowing function examination according to claim 1, characterized in that: Based on the size of the hyoid bone-thyroid cartilage distance in the hyoid bone-thyroid cartilage distance set, identifying the hyoid bone-thyroid cartilage distance shortening rate to obtain a target hyoid bone-thyroid cartilage distance shortening rate includes: Extracting the maximum value of the hyoid bone-thyroid cartilage distance and the minimum value of the hyoid bone-thyroid cartilage distance from the hyoid bone-thyroid cartilage distance set; The difference between the maximum hyoid bone-thyroid cartilage distance and the minimum hyoid bone-thyroid cartilage distance is calculated, and the ratio of the calculated result to the maximum hyoid bone-thyroid cartilage distance is used as the target hyoid bone-thyroid cartilage distance shortening rate.

Images